CONSIDERATIONS FOR FUTURE IGS RECEIVERS

TODD HUMPHREYS, LARRY YOUNG, AND THOMAS PANY

Abstract. Future IGS receivers are considered against the backdrop of GNSS signal modernizationand the IGS’s goal of further improving the accuracy of its products. The purpose of this paper is toprovide IGS members with a guide to making decisions about GNSS receivers. Modernized GNSS signalsare analyzed with a view toward IGS applications. A schedule for minimum IGS receiver requirementsis proposed. Features of idealized conceptual receivers are discussed. The prospects for standard com-mercial receivers and for software-defined GNSS receivers are examined. Recommendations are givenfor how the IGS should proceed in order to maximally benefit from the transformation in GNSS thatwill occur over the next decade.

1. Introduction

There are two reasons why it makes sense for the IGS to study GNSS receivers that will be integratedinto its network in the coming years. First, the new GNSS signals that will come on line over the nextdecade will render current IGS receivers obsolete, so it is prudent to examine receiver options goingforward. Second, the push to improve the accuracy of IGS products beyond current limits demandsgreater accuracy in the models used to describe receiver measurements. As a result, the IGS mustdemand from vendors more transparency into receiver firmware or adoption of user-specified algorithms.

This paper considers future IGS receivers from four different points of view. Section 2 looks atmodernized GNSS signals and their benefits for the IGS. Section 3 surveys the range of expected receivercapability. Section 4 considers current and future commercial geodetic-quality receivers. Section 5considers software GNSS receivers as an alternative to less reconfigurable traditional receivers. Section6 lays out the authors’ recommendations to the IGS.

2. Signals and Performance

GPS modernization is underway. Six signals are currently being broadcast from modernized GPSsatellites; a seventh signal is scheduled for on-orbit transmission before the end of 2008. Of the sixcurrent signals, two are the new military signals, M1 and M2, which cannot be tracked by unauthorizedreceivers. The other four are the C/A signal at L1 (1575.42 MHz) and the L2C signal at L2 (1227.6MHz), which can be tracked using open codes, and the two encrypted P(Y) signals transmitted at bothL1 and L2, which can be tracked by unauthorized receivers only if the receivers employ codeless orsemicodeless correlation techniques. The seventh signal, a broadband civil signal, will be broadcast atL5 (1176.45 MHz). Another civil signal, L1C at L1, will be available with the first GPS III satellites.

The rollout schedule shown in Fig. 1 reflects an optimistic estimate of L2C and L5 availability. The18 L2C-capable satellites already on orbit or manifested for launch, of which 10 are also L5-capable,offer exciting near-term opportunities to improve IGS products.

The GLONASS constellation transmits three signals roughly corresponding to GPS C/A, P(Y) (L1),and P(Y) (L2), although at distinct carrier frequencies somewhat displaced from L1 and L2. GLONASS

Todd Humphreys is with the Sibley School of Mechanical and Aerospace Engineering, Cornell University,Email:(teh25@cornell.edu).

Larry Young is with NASA’s Jet Propulsion Laboratory, Email:(Lawrence.E.Young@jpl.nasa.gov).Thomas Pany is with University FAF Munich, Email:(thomas.pany@unibw.de).2008 IGS Workshop, Miami Beach, FL.

1

(current and manifested)

IIR−M(current)6

IIF

2018

06 07 08 09 10 11 12

Year

13

Figure 1. Optimistic schedule for rollout of L2C-capable GPS satellites. Block IIFsatellites will also broadcast the new L5 civilian signal.

signals are displaced in frequency for different satellites, which causes difficulties for high-precision usersbecause instrumental delays and phase shifts are not common among satellites tracked by the samereceiver owing to the distinct instrumental effects at each satellite frequency.

When Galileo begins transmitting, a rich set of at least six additional signals will be available at ornear L1, L5 (called E5a and E5b), and E6 (1278.75 MHz).

What makes for good signals for science applications?

2.1. Carrier Frequency.

2.1.1. Minimizing Errors in the Iono-Free Linear Combination. The (first-order) ionosphere-free linearcombination of pseudorange measurements ρ1 and ρ2 from signals at frequencies f1 and f2 is given by

ρIF =[1 +

f22

f21 − f2

2

]ρ1 −

[f22

f21 − f2

2

]ρ2

To reduce the magnification of measurement errors in ρ1 and ρ2 by this linear combination, the factorf22 /(f2

1 − f22 ) should be made as small as possible. For a dual-frequency combination of current and

planned GNSS carrier frequencies, this requirement suggests the selection of L1 and L5/E5a.

2.1.2. Trilaning. The use of combinations of carrier phase observables to estimate carrier ambiguitiesshould be considered. In particular, by using three carriers, a new technique called trilaning becomespossible. If it assumed that a triple-frequency selection must include L1, L5, and an intermediatefrequency, then the proposed frequency for Galileo E6 is near optimum for trilaning.

To see this, suppose signals from L1, E6, and L5 are available so that widelane observables from L1-E6 and E6-L5 phases can be formed. Let these widelane observables be compared with those obtainedusing L1, L2, and L5 to see what is gained by using E6. Assume double-differenced measurements inthis discussion.

The first step is to use pseudoranges to resolve the widelane ambiguities. If the pseudoranges areassigned 10 cm errors, and carrier phase measurements have 1 mm error, widelane ambiguities cansuccessfully be estimated to < 1/6 cycle for both the L1/L2/L5 and L1/E6/L5 cases. (A 1/6 cyclecriterion is used because it yields a 3-sigma probability of selecting the correct integer ambiguity.)

The next step is to select two widelane observables and form the ionosphere-free combination. Thereis a multiplier between the error in individual phase measurements and the ionosphere-free range formedfrom three carrier phase measurements, resulting in the overall range error ∆ρ given in Table 1 for theL1/L2/L5 and L1/E6/L5 cases.

2

Table 1. Ionosphere-free trilane error ∆ρ for three trilane options

Carrier Frequency (×10.23 MHz)

L1 L5 L2, E6, or CS ∆ρ (cm)

154 115 120 (L2) 11.0154 115 125 (E6) 6.7154 115 500 (C-band) 0.3

2.1.3. C-band GNSS Signal. At least one of us favors investigation of a GNSS signal at a frequencymuch higher than those currently in use, near 5115 MHz for example. A signal at this frequency couldbe used to form very precise carrier phase observables in a regime where the ionospheric effect is muchsmaller. Smart, actively steered arrays could be built that are close to the size of today’s L-bandhemispherical antennas but would allow multiple beams toward the satellites and null forming towardmultipath sources.

2.2. BPSK Code Modulation.

2.2.1. Chipping Rate. There are two main advantages to higher chipping rates:(1) For a given ratio of signal bandwidth to chipping rate, the errors due to receiver (thermal)

noise are inversely proportional to the chipping rate. Figure 2 compares the ranging precisionof modernized GPS signals in the presence of white receiver noise as a function of the L1C/A carrier-to-noise ratio (C/N0). Typical L1 C/A C/N0 values for elevation angles above 10degrees range from 37 to 53 dB-Hz. Signal power levels consistent with Block IIR-M and IIFGPS satellites are assumed, i.e., relative to L1 C/A, L2C is 1.5 dB below, L1 P(Y) and L2P(Y) are 3 dB below, and L5 (either I5 or Q5) is 3.6 dB above. A squaring loss commensuratewith the near-optimal linear approximation to the MAP carrier recovery technique introducedin [1] is assumed (modified to account for the increased L2 P(Y) signal power transmitted bythe modernized GPS satellites). The squaring loss is reflected in the relatively steeper slope ofthe P(Y) curve in Fig. 2.

The benefit of the higher chipping rates of the P(Y) and L5 signals —ten times that of theL1 C/A and L2C signals—is obvious. Even when weakened by the squaring loss due to lackof knowledge of the W-bits, semicodeless tracking of the P(Y) signals on L1 and L2 results insmaller ranging errors than for L1 C/A and L2C above an L1 C/A C/N0 of approximately 47dB-Hz. Of course, the L5 signal, which combines higher power with a faster chipping rate, hasthe best ranging precision of the group, delivering an approximate 5-fold improvement over L1C/A and L2C.

(2) Multipath is a more complex issue. For multipath sources whose additional delay is less thanabout half the lag spacing (which is usually half a chip), the multipath error is the same for allchipping rates, as shown in Fig. 3. The strongest multipath signals come from nearby sources atIGS sites, since scattering losses are proportional to the distance squared. Furthermore, distantmultipath signals fluctuate more rapidly so that the resulting errors average down quicker thanerrors from nearby sources. It is known that nearby sources can add a systematic signaturealiasing into the local vertical and tropospheric estimates, but less is known about the effects ofdistant sources on estimated parameters at IGS sites. We recommend that studies be performedto determine whether multipath from signals delayed by more than 40 ns (approximately 12 m)contributes significantly to errors in the estimated parameters at IGS sites.

2.2.2. Pilot Signals. Many of the new GPS and Galileo codes will have a “pilot” component. Thesesignals carry no data bits, and so coherent integrations can be made over long intervals to allow easier

3

30 35 40 45 50 5510

−3

10−2

10−1

100

101

L1 C/A C/N0 value (dB−Hz)

Ran

ging

pre

cisi

on (

1−σ, m

) L2CL1 C/A

L1 or L2 P(Y)

L5 (I5 or Q5)

Figure 2. Ranging precision of modernized GPS signals in the presence of white re-ceiver noise as a function of the L1 C/A carrier-to-noise ratio (C/N0). Signal powerlevels consistent with Block IIR-M and IIF GPS satellites are assumed, i.e., relative toL1 C/A, L2C is 1.5 dB below, L1 P(Y) and L2 P(Y) are 3 dB below, and L5 (either I5or Q5) is 3.6 dB above.

97.8 ns E-L lag, -10 dB MP

-600

-400

-200

0

200

400

600

0 10 20 30 40 50 60 70 80 90 100

MP delay (ns)

C/A err (cm)

P41

Figure 3. C1 and P1 multipath error comparison assuming an early-minus-late corre-lator spacing of 97.8 ns and a multipath component whose power is 10 dB below that ofthe direct component. Multipath error is expressed in cm.

4

P code and BOC(10,5), -10 dB MP, 32 ns E-L lag spacing

-200

-150

-100

-50

0

50

100

150

200

0 10 20 30 40 50 60 70 80 90 100

MP delay (ns)

P1 multipath

BOC(10,5)

Figure 4. P-code and BOC(10,5) multipath error comparison assuming an early-minus-late correlator spacing of 32 ns and a multipath component whose power is 10 dB belowthat of the direct component. Multipath error is expressed in cm.

acquisition of weak signals. Moreover, because pilot signals allow full-cycle carrier tracking, whereas bi-phase modulated channels allow only half-cycle tracking, the tracking precision required to avoid carrierslips in the pilot tone tracking loop is reduced by about 6 dB. This translates into a 6-dB reduction inthe tracking threshold, which will make tracking through weak signals and through ionosphere-inducedscintillation more robust.

2.2.3. Coded vs. Semicodeless Tracking. Use of semicodeless processing has given science users accessto dual frequency signals in the past. Now that civil signals are beginning to be available at L2 aswell as at L1, and will soon be available at L5, one can consider the relative advantages of coded vssemicodeless tracking.

(1) Coded tracking has the advantage of higher signal-to-noise ratio (SNR) for equivalent receivedC/N0, allowing better acquisition and tracking in challenging conditions.

(2) At high C/N0 values, semicodeless tracking of higher rate codes can yield smaller ranging errorsdue to receiver noise than coded tracking of lower rate codes (cf. Fig. 2).

2.3. BOC and Multiplexed BOC Signals. These new signals will be exciting to exploit, obtainingthe advantages of high chipping rate BPSK codes (better precision and lower multipath for long mul-tipath delays) by the use of two or more subcarriers. Figures 4 and 5 compare P-code and BOC(10,5)code and carrier multipath. In addition, we anticipate that clever people will find unexpected ways toexploit the unique observables that BOC and MBOC codes produce.

3. Characteristics of Future IGS Receivers

This section considers a range of GNSS receiver capability, extending from the minimum that the IGScurrently allows to the maximum imaginable that still lies within the reach of practical implementation.As aids to discussing the upper limits of practical capability, two conceptual receivers are introduced—the Super and Ultra receivers.

5

Carrier MP

-1.5

-1

-0.5

0

0.5

1

1.5

0 10 20 30 40 50 60 70 80 90 100

MP delay (ns)

BOC(10,5) carrier

P1 Carrier MP

Figure 5. P-code and BOC(10,5) carrier multipath error comparison assuming anearly-minus-late correlator spacing of 32 ns and a multipath component whose power is10 dB below that of the direct component. Multipath error is expressed in cm.

3.1. Minimum Receiver Requirements. Current (June, 2008) minimum required observables forIGS receivers are summarized in Table 2 in terms of both the RINEX 2.11 and 3.00 standards. Inthis table, ‘or’ is meant to be inclusive; e.g., C1 or P1 implies at least one of C1 and P1. It is furtherassumed that in the absence of antispoofing, all receivers track P-based observables directly; i.e., viaautocorrelation with the known P-code. For convenience, observation code conventions for the RINEX3.00 standard are given in Fig. 6.

Table 2. Current minimum required observables for IGS receivers

RINEX 2.11 RINEX 3.00

L1 L1C or L1P or L1W or L1Y or L1NL2 L2C or L2D or L2S or L2L or L2X or L2P or L2W or L2Y or L2M or L2NP2 C2D or C2P or C2W or C2YC1 or P1 C1C or C1P or C1W or C1Y

Over the next decade or so, these minimum requirements must be updated to ensure that the IGSfulfills its mission as the “premier source of the highest-quality GNSS related standards (conventions),data, and products” (IGS Strategic Plan 2008-2012). Not only will the IGS want to exploit new GNSSsignals as they come on line, but it must also adapt to the recent announcement that the U.S. Air Forceintends to discontinue codeless and semicodeless access to the encrypted P(Y) signals on L1 and L2 byapproximately 2020 (see IGS Mail 5774).

In proposing updates to the minimum receiver requirements, the IGS must temper its desire toremain current with a recognition that abrupt, sweeping requirement changes are not appropriate fora loose volunteer federation such as the IGS. It would not be prudent, for example, to require that allreceivers in the IGS network be L2C-capable by 2009. The following proposed requirements changesrepresent, in the authors’ judgment, a sensible requirements update schedule for the IGS. Accordingto this schedule, minimum requirements changes are linked to specific events, not to calendar dates.Changes listed under each event are relative to the current requirements at the time the event occurs.In other words, after each event all the previous requirements apply plus those specified for the event.Events A1, A2, and A3 will occur in that order, but the full chronology of events Ax through C is not

6

Figure 6. RINEX 3.00 observation codes.

7

currently known. Requirements specifications are expressed in terms of the RINEX 3.00 observationcodes (cf. Fig. 6).

Event A1: 8 or more L2C-capable GPS satellites (∼2009). The following requirementschanges apply to receivers newly incorporated into the IGS network after this event:(1) Receivers must measure at least one of GPS L2S, L2L, and L2X.(2) Receivers must measure at least one of GPS C2S, C2L and C2X.(3) Receivers must have the capability of being slaved to an external frequency reference.

Event A2: 8 or more L5-capable GPS satellites (∼2012). The following requirementschanges apply to receivers newly incorporated into the IGS network after this event:(1) Receivers are no longer required to measure any of GPS C2D, C2P, C2W, and C2Y.(2) Receivers must measure at least one of GPS L5I, L5Q, and L5X.(3) Receivers must measure at least one of GPS C5I, C5Q, and C5X.

Event A3: Discontinuation of codeless or semicodeless access to P(Y) signals (∼2020).Following this event, the requirements changes applied to newly incorporated receivers afterEvents 1 and 2 now apply to all receivers in the IGS network.

Event B: 8 or more Galileo satellites. The following requirements changes apply to receiversnewly incorporated into the IGS network after this event:(1) Receivers must measure at least one of Galileo L1A, L1B, L1C, L1X, and L1Z.(2) Receivers must measure at least one of Galileo C1A, C1B, C1C, C1X, and C1Z.(3) Receivers must measure at least one of Galileo L5I, L5Q, L5X, L8I, L8Q, and L8X.(4) Receivers must measure at least one of C5I, C5Q, and C5X, C8I, C8Q, and C8X.

Event C: 8 or more dual-frequency CDMA GLONASS satellites. The following require-ments changes apply to receivers newly incorporated into the IGS network after this event:(1) Receivers must measure at least one of GLONASS L1C and L2P.(2) Receivers must measure at least one of GLONASS C1C and C1P.(3) Receivers must measure at least one of GLONASS L2C and L2P.(4) Receivers must measure at least one of GLONASS C2C and C2P.

3.2. The Super Receiver. The Super Receiver is a conceptual GNSS receiver with characteristicsthat are desirable to the IGS. Its features could be implemented and its observables processed withcurrent technology.

The Super Receiver• tracks all open signals on all healthy GNSS satellites;• tracks encrypted signals whenever there is no open signal on the same carrier or whenever

their pseudorange precision in the presence of white (receiver) noise is better than that ofun-encrypted signals on the same carrier;

• is compliant with the latest RINEX standard;• is completely user reconfigurable via the Internet, from correlations to tracking loops to navi-

gation solution;• implements internal cycle slip mitigation and detection;• produces quadrature correlation products (Is and Qs) for all tracked signals at up to 50 Hz;• produces observables that can be accessed via the Internet;• is inexpensive.

3.3. The Ultra Receiver. The Ultra Receiver is a conceptual GNSS receiver whose products—whilecurrently impractical to store and process—would represent the highest quality, most useful, most stableGNSS observables imaginable for the IGS.

8

4

The Ultra Receiver

Digital Storage Rx

MassStorage

RF Front-End

ReferenceOscillator

ADC

SampleClock

Figure 7. A digital storage GNSS receiver: the single-channel component of an Ultra Receiver.

The Ultra Receiver is composed of a bank of digital storage receivers like the one shown in Fig. 7.Each digital storage receiver captures the entire spectral content of a particular GNSS signal by wide-band filtering (and perhaps downconverting) the signal, digitizing it at a sufficiently high samplingrate, and storing the samples to disk. Two-bit sampling quantization would be adequate for foreseeableapplications.

The high-sampling-rate data stored to disk could be immediately processed for near-real-time oper-ation; but, more importantly, it could be reprocessed at any later date with the best signal processingstrategy of the day. Such a reanalysis of archived GNSS data would have the potential to eliminatefrom IGS products all effects due to receiver idiosyncrasies and upgrades. The stability of IGS productswould then be limited only by the evolution of the network itself (i.e., by the number of sites and theirlocations).

Of course, a continuously-operating Ultra Receiver is currently impractical for two reasons: (1) datastorage requirements would be stupendous and, (2) reprocessing would require large computer clustersor long execution times. Consider, for example, a sub-Ultra Receiver that captures only the GPS L1C/A signal. For adequate phase stability, a wide passband filter—say, 8 MHz—would be required. Toavoid aliasing effects, the bandpass sampling theorem would require a sampling rate of at least 16 Msps.Such a staggering data stream would fill a 1 Tera-byte hard drive (roughly the largest consumer-gradehard drive available today) in less than a week. Reprocessing the stored data using a single CPU couldbe done presently at about 10 times real time. This would mean that reprocessing 10 years of data froma single station to produce standard receiver observables would require one year on a single computer.

Despite being currently impractical, a continuously-operating Ultra Receiver is not far off, and theIGS would be wise to aspire to it in the coming decades.

4. Commercial Geodetic-Quality Receivers

This section examines current commercially-available geodetic-quality GNSS receivers from the IGSperspective. Commercial receivers with the reconfigurability and other features of the Super Receiverdescribed above would be of great interest to the IGS. Short of this, the IGS expects at least high-quality dual-frequency range and phase measurements and—so that measurements can be modeledcorrectly—transparency into the recipe used by each receiver to make such measurements.

To probe the suitability of current commercial receivers for IGS applications, a questionnaire wassent to four GNSS receiver vendors: Septentrio, Trimble, Leica, and Topcon. All four vendors made aserious effort to provide detailed responses to the questionnaire.

9

These four vendors were chosen because their products are currently the ones most commonly chosento replace ailing receivers across the IGS network. The following subsection, based on responses tothe questionnaire, interviews with IGS members, and the authors’ judgment, gives the outlook forcommercial receivers over the next several years. Thereafter, the full questionnaire and responses arepresented.

4.1. Outlook.

4.1.1. Favorable Aspects.

Signal tracking capability: There has been some worry in the past that receiver vendors wouldnot equip their receivers to track modern GNSS signals until a full complement of modern GNSSsatellites had been launched. This is not the case. All vendors who responded to the survey offeror will soon offer L2C and L5 capability, dual-frequency GLONASS tracking, and have plansto offer Galileo capability by the time there are a useful number of Galileo satellites (probablyaround 6) on orbit.

Internet readiness: Vendors are offering receivers that support a web page for easy remotereceiver monitoring—and in come cases reconfiguration—and an FTP server for data retrieval.

Reconfigurability: Septentrio and Leica allow users to reconfigure phase-locked loop (PLL)bandwidths.

Data quality: All vendors appear to track GNSS signals with algorithms that are near op-timal at high C/N0 values. The code noise, phase noise. and cycle slipping performanceof current receivers—even for semi-codeless recovery of the L2 carrier and P(Y)-based codemeasurements—appears to be very good (for example, compare the admirable performance ofthe Septentrio PolaRx2 with the BlackJack “Gold Standard” receiver in [2]).

Platform: Current commercial receiver hardware has proven to be generally rugged and reliable.

4.1.2. Unfavorable Aspects.

Measurement definition: Precision GNSS users must be able to accurately model GNSS re-ceiver measurements in estimation algorithms used to process the measurements. For example,the mean, variance, and time correlation of errors associated with a particular measurementmust be accounted for in a measurement model. Such a model is easy to construct if the exactrecipe for how measurements are made is known. At the very least, precision users need toknow the effective measurement interval to which a given phase, pseudorange, or signal powermeasurement corresponds. If measurements are made by simply sampling the phase, code,and signal power tracking loops, then the loops’ bandwidths can be used to infer an effectivemeasurement interval.

Trimble, one of the largest vendors of geodetic-quality receivers, was unwilling to discloseessential details about how their receivers’ phase, pseudorange, and signal power measurementsare made. Other vendors either disclosed tracking loop bandwidths or have made these config-urable.

Even better than knowing tracking loop bandwidths would be persuading vendors to adopta standard measurement technique, such as the one described in JPL’s recently expired patent“Digital signal processor and processing method for GPS receivers” (U.S. patent 4,821,294).Apparently, NovAtel has adopted the JPL technique for use in their WAAS reference stationreceivers. Other vendors may follow suit considering the excellent code and carrier trackingproperties of the JPL BlackJack receiver, which employs the technique (see [2] for an evaluationof BlackJack code and carrier tracking performance). Such standardization would lead to bettermeasurement models and facilitate combining measurements from different receivers.

10

Proprietary output formats: The Trimble NetR5 receiver outputs data in a proprietary binaryformat. The IGS finally obtained a conversion utility from Trimble two years after an initialrequest.

Signal-to-noise ratio reporting: Most receivers now output a measure of SNR in dB units,which is good, but the quality of the SNR measurement is not uniform among vendors oramong signals on a particular receiver. For example, it was reported at the 2008 IGS Workshopthat the Trimble NetRS receiver produces noisy L1 C/A SNR measurements but clean L2CSNR measurements. It has also been reported in the literature that the Septentrio PolaRx2 L2P-code SNR measurement is not accurate [2].

Limited reconfigurability: Because of their limited reconfigurability, receivers from commercialvendors do not support some observables and tracking techniques that may be useful to theIGS. For example, as was advanced by Thomas Pany in the 2008 IGS Conference, the IGSwould benefit from receivers capable of outputting the complex baseband signal (Is and Qs)associated with each distinct tracked signal. A 50-Hz output rate would be adequate for mostapplications. Among other uses, such data from two GNSS receivers could be combined to docorrelation-level double differencing, a side effect of which is the wipeoff of navigation data bitsfrom the double differences. All GNSS receivers produce this measurement internally, but fewcommercial receivers make it available to the user.

Consider another example where receiver reconfigurability would benefit the IGS. With theadvent of L2C, one could imagine L2C-aided semicodeless tracking of the P(Y) code on L2. Thecurrent technique is to use the L1 C/A carrier for aiding, but because the L1 and L2 carrierphases diverge during strong ionospheric events, L2C aiding would lead to a more robust P(Y)-based pseudorange measurement on L2. Apparently, no vendors are currently considering sucha tracking strategy. Other exotic tracking techniques that would be useful for tracking weakand scintillating signals are likewise unavailable on commercial platforms.

4.2. Responses to Questionnaire. The four GNSS receiver vendors to whom the questionnaire wassent recommended the following receivers for use in the IGS network. Prices are approximate list pricesfor receivers without antennas. Leica, Trimble, and Topcon indicated that reduced prices are availablefor educational or non-profit organizations.

Septentrio: PolaRx3 ($9,750)Trimble: NetR5 ($20,000) or NetRS ($16,000)Leica: GRX1200 GG Pro ($22,000)Topcon: Net-G3 ($18,500)

Responses in the questionnaire should be understood to refer respectively to these receivers unlessotherwise noted. The questionnaire and responses follow:

Timing considerations. In order to combine measurements from receivers made by different vendors,the IGS needs to know how (1) the measurement epoch, and (2) the measurement interval are definedfor each receiver.

(1) The RINEX 2 definition of the measurement epoch is “The time of the measurement is thereceiver time of the received signals. It is identical for the phase and range measurements andis identical for all satellites observed at that epoch. It is expressed in GPS time (not UniversalTime).” Some receivers define the measurement epoch to coincide with the C/A code boundaryof one of the signals being tracked. Other receivers define the epoch to coincide with integerseconds per the receiver clock. Still other receivers define the phase and pseudorange measure-ment times separately (contra the RINEX definition above). Please describe your definition ofmeasurement time.

11

Septentrio: Septentrio receivers define the epoch to coincide with integer seconds per the re-ceiver clock (or fraction thereof if the measurement rate is higher than 1 Hz). The definitionis fully compliant with the RINEX standard.

Trimble: Trimble’s receivers are fully compliant to the RINEX 2(and 3) standardLeica: The Leica System1200 receiver engine is fully compliant to the RINEX standard.Topcon: Topcon receivers are fully compliant to the RINEX standard. Epoch time is referenced

to receiver clock and is always within 0.5 milliseconds from GPS time. Clock correction(receiver clock offset) observable is available in native format and can be output in RINEX.

(2) The measurement interval is the interval over which the data are observed to produce phase,pseudorange, or signal power measurements. Some receivers simply sample the PLL phase es-timate and the DLL code offset estimate to produce phase and pseudorange measurements,respectively. In this case, the measurement interval is given by the time constants of these twotracking loops (which go inversely as the loops’ bandwidth). Other receivers use a measurementinterval that is independent of the tracking loop bandwidths. For example, the JPL BlackJackreceiver uses a known number of 0.02-second data intervals symmetrically arranged before andafter the measurement time to obtain data for a given pseudorange (or phase or C/N0) mea-surement, and there is no correlation of noise between adjacent measurements. Please describeyour definition of measurement interval and its position relative to the measurement time.Septentrio: Septentrio receivers sample the continuous PLL phase, DLL code and C/N0 es-

timates at the measurement epoch. There is no additional fit over internal higher-ratemeasurements.

Trimble: PLL characteristics and algorithms are proprietary and of significant commercial value.Trimble does not disclose this information.

Leica: Leica System1200 receivers sample the continuous PLL phase, DLL code and C/N0

estimates at the measurement epoch.Topcon: Topcon receivers sample the continuous PLL phase and DLL code at the measurement

epoch. C/N0 estimates are obtained at 10 Hz.

Phaselock loop characteristics. To anticipate how well a receiver might withstand the effects ofionospheric scintillation, it is useful to know the phaselock loop update (pre-detection) interval (e.g.,1ms, 10ms, 20ms), its order (e.g., 1st, 2nd, or 3rd), and its bandwidth (e.g., 10 Hz).

Septentrio: The pre-detection interval and bandwidth is user selectable, with defaults being 10ms and10 Hz respectively. A 3rd order PLL is used.

Trimble: PLL characteristics and algorithms are proprietary and of significant commercial value. Trim-ble does not disclose this information.

Leica: The Leica System1200 receivers pre-detection interval is not user selectable, however it auto-matically adjusts based on the PLL bandwidth selected by the user. The bandwidth is userselectable in a special version of the firmware available from Leica. The default is 15Hz. A 3rdorder PLL is used.

Topcon: The pre-detection interval is not user-selectable. The bandwidth and the order are user-selectable. The default bandwidth is 25Hz and can be changed from 2Hz to 50Hz. A 3rd orderPLL is used by default and can be changed to 2nd order.

Reference oscillators. The phase stability of a receiver depends on the stability of its oscillator.Please indicate what type of internal reference oscillator your receiver uses, whether TSXO, TCXO,OCXO.

Septentrio: Internal TCXO, which can be bypassed by feeding an external frequency reference.Trimble: TCXOLeica: The Leica System1200 receiver onboard TCXO can be slaved to an external frequency reference.

12

Topcon: Net-G3 receiver uses internal TCXO. It can be slaved to an external frequency reference (5MHzto 20MHz).

Features. The IGS is interested in procuring GNSS receivers that are well adapted to collecting datauseful for ionospheric, tropospheric, and geodetic science applications. Please indicate whether yourreceiver currently supports the following features, or whether these features could be added as part offuture firmware upgrades or receiver models.

(1) Reconfiguration of tracking loops: Is it possible to change the bandwidth of the receiver’s PLLand DLL? What about changing the loops’ update intervals? Does this require a firmwareupgrade? Can these loops be reconfigured remotely; i.e., over the network?Septentrio: Bandwidth and pre-detection time can be changed by a single user-command. No

firmware upgrade is needed. The command can be sent remotely like any other commandof the receiver.

Trimble: The tracking loop bandwidths and update intervals are not configurable.Leica: The Leica System1200 receiver’s bandwidth for the PLL and DLL can be changed

through the firmware (a special firmware can be provided to do this). The pre-detectiontime cannot be changes through a user command (a firmware change would be required).

Topcon: Tracking loops’ update intervals are not configurable. The bandwidth can be changedremotely and this does not require a firmware upgrade. The bandwidth of PLL can bechanged from 2Hz to 50Hz, DLL from 0.1Hz to 50Hz.

(2) Can firmware upgrades be delivered over the network to a remote receiver?Septentrio: Yes in future firmware upgrades.Trimble: Firmware upgrades for the Trimble NetRS, NetR3, and NetR5 can be delivered and

implemented via Ethernet.Leica: Yes, either via the web interface or Leica GNSS Spider or via our proprietary LB2/OWI

interface.Topcon: Yes.

(3) At what frequency can the receiver measure the amplitude (or C/N0) and phase of each signalbeing tracked (e.g., 10 Hz, 20 Hz, 50 Hz). If less than 50 Hz, can the rate be increased to 50Hz via a firmware upgrade or otherwise?Septentrio: 50 Hz.Trimble: C/N0 is measured at 10 Hz in the NetR5. Phase measurements occur at a maximum

rate of 20 Hz.Leica: The C/No update rate is 20Hz. It would require a special version of the firmware to

support a higher rate such as 50Hz.Topcon: Net-G3 receiver can measure the phase of each signal at 20Hz. The C/N0 update rate

equals to 10Hz. It requires a special version of firmware to support 50Hz update rate.Future Topcon receivers will have 100Hz update rate option.

(4) Does your receiver support L2C tracking? If so, does it track both the L2CM and L2CL codes?If the L2CL code is tracked, is this done with a non-squaring PLL?Septentrio: Yes, L2C is tracked in the optimal mode (i.e. tracking L2CL with a non-squaring

PLL).Trimble: The NetR5 supports L2C tracking. It is user configurable to provide L2CM, L2CL,

or L2CM+L2CL.Leica: Yes, the Leica System1200 receiver supports L2C tracking. We track both L2CM and

L2CL codes. In the current firmware L2C tracking is aided by L1 tracking. Future receiverswill have independent L2C tracking.

Topcon: Yes, Net-G3 receiver supports L2C tracking (both L2CM and L2CL). In the futureversion of firmware it can be configured to track either L2CM, or L2CL (or both of them).

(5) Does your receiver support L5 tracking?13

Septentrio: Yes, it can be configured in L1/L2 mode, or L1/L5 mode. The PolaRx3 does notsupport concurrent tracking of L2 and L5. Future receiver models (PolaRx4) will.

Trimble: The NetR5 supports L5 tracking.Leica: Currently the System1200 does not support L5. The receiver hardware will be available

in 2009. All existing System1200 receivers may then be upgraded to support L5 and Galileowith a board exchange.

Topcon: Net-G3 receiver hardware is capable of tracking of L5 signal. A firmware upgrade isrequired for getting L5 tracked.

(6) Does your receiver support P-code tracking on L1 and L2? Does it produce P-code-basedpseudorange measurements on both L1 and L2? Do you intend to support P-code tracking evenafter L5 becomes fully available?Septentrio: Tracking is performed on P1 and P2, but only P-L2 measurements are output. We

have no plan to discontinue the support of P-code tracking.Trimble: The NetR5 does not support full P-code tracking by fully decrypting Y code signals.

P-code based measurements are only available for L2. There are no plans to reduce thedata available from Trimble’s GNSS Infrastructure receivers as additional signals becomeavailable. Trimble designs the receiver to track and stream all available signals and providesthe end user with the option to configure reduced tracking or streaming to suit theirparticular application.

Leica: The Leica System1200 engine tracks P-code on L2 only. We have no plan to discontinuethe support of P-code tracking.

Topcon: Topcon receivers support P-code tracking on L1 and L2, thus carrier phases and pseu-doranges on both L1 and L2 bands can be measured. There are no plans to reduce thisfunctionality in future Topcon receivers.

(7) Which Galileo signals do you intend to support, and when do you anticipate that your receiverswill be capable of doing so?Septentrio: Current solutions: PolaRx3: E5a, L1B and L1C, all in view.

GeNeRx: all types of GIOVE and Galileo signals, limited to 4 satellites. Future: we areplanning to support full E5 and possibly E6 for all satellites in view, but this capabilitymay not be available until 2010, or thereafter.

Trimble: Which signals we intend to support is proprietary. This will depend on the commercialavailability of Galileo signals.

Leica: The next generation of measurement engine will support L1/L2/L5 GPS + L1/L2GLONASS + E1/E5a/E5b/Alt-BOC. The receiver hardware will be available in 2009.All existing System1200 receivers may then be upgraded to support Galileo and L5 with aboard exchange. A firmware update will be required to enable the Galileo tracking whenthe constellation is large enough (e.g. 2 satellites) to justify it.

Topcon: Net-G3 receiver can already track E1/E5a signals from GIOVE satellite with a specialversion of firmware. Future Topcon receivers will track all Galileo signals.

(8) Which GLONASS signals do you intend to support, and when do you anticipate that yourreceivers will be capable of doing so?Septentrio: PolaRx3 currently supports L1-C/A, L2-P. We also intend to incorporate L2-CA.Trimble: NetR5 currently supports GLONASS L1 C/A & P, L2 C/A & P.Leica: The Leica System1200 engine supports L1-CA and L2-P. We can use L2-CA with a

firmware change.Topcon: All Topcon receivers can track GLONASS signals. Net-G3 receiver currently supports

L1 C/A & P, L2 C/A & P GLONASS signals.(9) Are the contents of your receiver’s binary output completely specified in documentation that

the user can obtain? (This is required for data conversion to the latest RINEX format).

14

Septentrio: Yes, and additionally a Binex (open source) format conversion software tool for ourreceivers is currently being created by NR Canada for IGS use.

Trimble: The NetR5 offers various binary outputs, both proprietary and publicly specified.Most RINEX format files can be produced from an RTCM3.0 data stream. Trimble haspreviously made proprietary message formats available to customers needing this informa-tion for their application but makes no promise that this is always possible.

Leica: Yes, we provide the proprietary LB2/OWI interface of our receivers to customers.Topcon: Yes. A software tool for conversion of TPS-files to RINEX is provided. TPS files can

also be converted to RINEX with current version of teqc from UNAVCO.(10) Can the code running on your baseband processor be obtained under a license? If so, what is

the approximate cost of the license?Septentrio: This option is available, dependent upon the user requirements and potential license

restrictions. Prices have not yet been established for this option.Trimble: Baseband processor code cannot be purchased from Trimble.Leica: No, not at this time.Topcon: No, not for current receivers.

5. Software GNSS Receivers

One of the long-term goals identified in the 2008-2012 IGS strategic plan is to “incorporate and inte-grate new systems, technologies, applications and changing user needs into IGS products and services.”In keeping with this goal, the authors believe the IGS should keep a close watch on the development ofsoftware GNSS receivers and possibly incorporate software receiver technology into future IGS referencestations.

Software GNSS receivers (also known as software-defined receivers) differ from traditional GNSSreceivers in that they do not depend on any special-purpose ASIC chips. Instead, a software receiverimplements all digital signal processing functions (i.e., those downstream of its analog RF front end)—from correlation to navigation solution—in software reprogrammable FPGAs, DSPs, or general-purposemicroprocessors. The advantages in flexibility that software receivers offer, coupled with ever-more-powerful processors, suggest that software receivers will have an expanding presence in future specialtyGNSS applications like those supported by the IGS.

The following section gives an overview of software GNSS receiver prospects from an IGS perspective.Thereafter, reports on the status of three significant software receiver development efforts are given.

5.1. Outlook.

5.1.1. Favorable Aspects.Reconfigurability: All signal processing downstream of digitization is completely reconfigurable

on a software GNSS receiver. Hence, software receivers can readily support non-standard ob-servables and tracking loops, internal cycle slip detection and mitigation, and other adaptationsinteresting to the IGS.

Transparency: The software receiver development effort at Cornell University has plans to licensethe complete receiver source code to academic and scientific organizations. The JPL receiver hasprecisely and openly defined measurement techniques. The commercial version of the UniversityFAF Munich receiver, the IFEN GmbH NavX(R)-NSR, offers sockets into which users can plugtheir own tracking loop code. All of these approaches allow users to accurately model thereceiver products.

Performance: Because software receivers can emulate the behavior of the special-purpose ASICsused in GNSS receivers—indeed, it is common industry practice to develop an FPGA-basedfunctional prototype of an ASIC before fabrication—the performance limits of software receiversare no different from those of traditional receivers.

15

5.1.2. Unfavorable Aspects.Lack of P(Y) tracking: Besides the JPL software receiver, which employs FPGAs to support

high-sample-rate correlations, no other software GNSS receiver of which the authors are aware iscurrently capable of codeless or semicodeless tracking of the P(Y) signal on L1 or L2. Given thatthis capability is expected of IGS receivers until Event A2 in the minimum requirements scheduleproposed in this paper (cf. Section 3.1), this represents a significant drawback. Of course,software receiver development efforts may include some kind of semicodeless P(Y) tracking infuture versions; however, to avoid intellectual property conflicts and to make best use of limitedcomputational resources, developers are less interested in the encrypted Y-code signals than inthe modernized GNSS signals.

Inadequate testing: For software receivers to gain a foothold in the IGS network or in anyother precision application they will have to undergo extensive comparison testing against morefamiliar traditional receivers. A comparison like the one conducted in [2] would be a useful pre-liminary test. Thereafter, several software receivers should be co-located with current receiversat IGS sites and their residuals, multipath performance, and day boundary discontinuities shouldbe examined just as for other IGS receivers.

5.1.3. Unknowns. Questions remain about how software receiver development will progress and how(and if) software receivers will be incorporated into the IGS network. One could think of these unknownsas unfavorable aspects, but they are more aptly categorized as simply unknowns.

Hardware platforms and software: Who will build the hardware platforms that support soft-ware receivers? Who will provide the software, support, and maintenance? In the most likelyscenario, a commercial provider will sell platforms to the IGS and provide software licenses andsupport.

One candidate, ASTRA LLC (www.astraspace.net) has plans to license the Cornell dual-frequency receiver code and build platforms to support it. ASTRA LLC may be willing tosell receiver platforms separately from the software license and maintenance package. In thiscase, the IGS may wish to obtain a software license—which would presumably be free foracademic and science users—directly from Cornell. However, Cornell would not provide softwaremaintenance or support. Hence, if the IGS chose this option, it would have to assume theresponsibility of tailoring the code for IGS applications and maintaining it. Code maintenancecould be centralized or could be distributed among the various IGS Analysis Centers.

Another vendor, IFEN GmbH (www.ifen.com), sells a commercial version of the UniversityFAF Munich receiver. IFEN GmbH sells its 10 MHz L1 front-end, the software receiver ex-ecutable, and the application programming interface (API) separately. Along with the APIcomes C/MATLAB source code, which could be the basis for future IGS software receiverdevelopment. A multi frequency front-end is in development.

Price: What will be the cost of commercial software receiver platforms, software licenses, andmaintenance contracts? Will per-unit costs be competitive with those of traditional receiversfrom established vendors? ASTRA LLC has plans to offer a dual-frequency (L1 C/A and L2C)platform for approximately $1,200. Software licensing and a maintenance contract will likely besold separately, with the per-unit cost of the software license and maintenance contract beingquite small (a few hundred dollars) for lots of 10 receivers. Approximate prices for the separatecomponents of the IFEN GmbH NavX(R)-NSR receiver are not currently known.

5.2. Status of Significant Software GNSS Receiver Development Efforts.

5.2.1. Cornell University. Cornell University has developed the GRID receiver (GNSS Receiver Imple-mentation on a DSP) with the goal of using tens or hundreds of GRID receivers to do dense-array-basedionospheric imaging. The GRID receiver, shown in Fig. 8 was formerly an L1 C/A-only receiver (see [3]

16

Figure 8. The Cornell University GRID (GNSS Receiver Implementation on a DSP) instrument.

for a writeup) but is currently undergoing a conversion to dual-frequency (L1 C/A and L2C) capabil-ity. The receiver is implemented on a Texas Instruments DSP and can operate as a stand-alone device.It includes advanced tracking loops specially designed for robustness in the presence of ionosphericscintillation.

The Cornell development team hopes to be awarded Phase II funding from an AFOSR STTR pro-gram. If awarded, the funding will allow the team to develop the GRID into a commercializable,internet-ready, fully-reprogrammable, dual-frequency, low-cost, software GNSS receiver.

5.2.2. University FAF Munich. Software GNSS receiver development at the University FAF Munichstarted in 2002. It has been funded by the German Armed Forces (BWB, AGeoBw), the DLR and ESA.The current receiver, a screenshot of which is shown in Fig. 9, is capable of tracking all-in-view civilGPS, SBAS and Galileo signals with a high accuracy on L1/L2 and L5. Processing of the encryptedP(Y)-code signal has not yet been done in real time. The approach pursued at the University FAFis based on the following key elements: a software architecture which supports multiple CPU coresto do signal tracking in parallel, a tracking scheme based on code continuous reference waveforms(CCRW) which allows multipath mitigation and is flexible with regards to the used sampling frequency,a dynamic sample rate reduction scheme which is employed when the received signal power is highand the contribution of thermal noise to the error budget is negligible. Most importantly, optimizedassembly language routines streamline the signal processing required for correlation and acquisition.

5.2.3. NASA’s Jet Propulsion Laboratory (JPL). As part of NASA’s Instrument Incubator Program,the GNSS receiver group at JPL has built a prototype instrument, TOGA (Time-shifted, Orthometric,GNSS Array), to address a variety of GNSS science needs (see Fig. 10). The TOGA receiver designfeatures several innovative capabilities:

(1) Multiple FPGAs to provide reprogrammable digital signal processing logic.(2) Buffer memory to store sampled data for either near-realtime onboard processing or processing

offline on one of JPL’s software receivers.(3) An electronically steered antenna (ESA) array which forms simultaneous beams in multiple

directions for L1, L2, and L5 frequencies to increase signal gain and suppress multipath.(4) A Linux operating system based science processor serves as experiment scheduler and data

post-processor to ease scientist-developed infusion of new algorithms.17

Figure 9. University FAF software receiver in GPS/Galileo mode.

The first TOGA engineering model receiver has been built and demonstrated by forming multiplebeams and, in post-processing, tracking L1 and L2 signals, L5 from a WAAS SBAS satellite, E1 fromGIOVE-A, and E1 and E5a from GIOVE-B.

6. Recommendations

(1) Study the effects of long-delay multipath by comparing co-located dual-frequency P(Y)-code-based measurements with dual-frequency C/A- and L2C-based measurements (cf. Section2.2.1).

(2) Adopt the minimum IGS receiver requirements schedule outlined in Section 3.1.(3) Request from commercial receiver vendors either (1) a detailed measurement description, or (2)

adoption of a standardized measurement technique (cf. Section 4.1.2).(4) Compare the performance of at least one software GNSS receiver against that of a traditional

receiver via signal simulator tests (such as those conducted in [2]) and via co-location withtraditional IGS receivers (cf. Section 5.1.2).

(5) Establish an IGS format for exchange of digitally-sampled IF data (i.e., the data stored to diskby digital storage receiver like the one schematized in Fig. 7).

(6) Form an IGS Software Receiver Working Group.

References

[1] K. T. Woo, “Optimum semi-codeless carrier phase tracking of L2,” in Proc. ION GPS 1999. Nashville, Tennessee:Institute of Navigation, 1999.

[2] O. Montenbruck, M. Garcia-Fernandez, and J. Williams, “Performance comparison of semicodeless GPS receivers forLEO satellites.” GPS Solutions, vol. 10, pp. 249 – 261, 2006.

[3] T. E. Humphreys, B. M. Ledvina, M. L. Psiaki, and P. M. Kintner, Jr., “GNSS receiver implementation on a DSP:Status, challenges, and prospects,” in Proceedings of ION GNSS 2006. Fort Worth, TX: Institute of Navigation, 2006.

18

Figure 10. JPL’s TOGA (Time-shifted, Orthometric, GNSS Array) instrument.

19